Deep Research

Other

Rigorous multi-source, fact-checked research producing a cited brief. Use for any high-stakes question where being wrong is costly — market sizing, company/person due diligence, investment targets, legal/regulatory, location feasibility. Fan out across sources, verify claims adversarially, cite everything.

Install

openclaw skills install @serkaion/serkaion-deep-research

Deep Research

For high-stakes questions where a wrong answer carries real cost. Depth and verification over speed.

Method

  1. Scope — Restate the question in one line. List 3–6 sub-questions that must be answered for a confident decision. If the question is vague, narrow it before researching.
  2. Fan out — Research each sub-question from multiple independent angles: web search, official/primary sources, registries and filings, reputable news, and any documents provided. One angle never suffices.
  3. Verify adversarially — For each material claim, find a second independent source and actively try to disprove it. Flag anything single-sourced, contested, or stale.
  4. Synthesize
    • Bottom line (3 sentences, decision-oriented)
    • Key findings — each with source + confidence (High / Med / Low)
    • What's uncertain or contested
    • Implications for the decision at hand
    • Recommended next action
  5. Cite every claim inline (source + date). No uncited assertions.

Rules

  • Never present a single-source claim as fact — always label confidence.
  • Separate fact from inference explicitly.
  • If you cannot verify something, say so — never fill gaps with plausible guesses.
  • Breadth first, then depth on the decision-critical items only.